Struct ConsensusBaseBuilder 

pub struct ConsensusBaseBuilder { /* private fields */ }

Builder for calling consensus at a single base position

Accumulates observations (base + quality) and uses a likelihood model to call the consensus base and quality.

Uses Kahan summation algorithm with SIMD vectorization to maintain numeric stability when accumulating log-likelihoods across many observations. The 4 DNA bases map perfectly to f64x4 (256-bit AVX2), enabling parallel computation of all 4 likelihoods.

Implementations

impl ConsensusBaseBuilder

pub fn new( error_rate_pre_umi: u8, error_rate_post_umi: u8, ) -> ConsensusBaseBuilder

Creates a new consensus base builder

Arguments

• error_rate_pre_umi - Phred-scaled error rate prior to UMI integration
• error_rate_post_umi - Phred-scaled error rate after UMI integration

pub fn reset(&mut self)

Resets the builder to process a new position

pub fn add(&mut self, base: u8, qual: u8)

Adds an observation (base + quality) to the consensus

Uses SIMD-vectorized Kahan summation for numeric stability when accumulating log-likelihoods. All 4 base likelihoods are updated in parallel using f64x4.

Arguments

• base - The observed base (A, C, G, T, or N)
• qual - The base quality score

pub fn call(&self) -> (u8, u8)

Calls the consensus base and quality

Returns (consensus_base, consensus_quality) Returns (N, 0) if no observations or multiple equally likely bases

Panics

Panics if there are observations but no maximum likelihood base is found (should not occur in practice since we check for ties).

pub fn contributions(&self) -> u16

Returns the total number of contributing observations

This is the sum of observations across all four bases

pub fn observations_for_base(&self, base: u8) -> u16

Returns the number of observations for a specific base

Arguments

• base - The base to query (A, C, G, or T)

Returns 0 for invalid bases

pub fn all_observations(&self) -> [u16; 4]

Returns the observations for all bases as [A, C, G, T]